I think this is the same anomaly as I noticed at [substr] anomaly or mine?.

I raised a perlbug (I can't find the ticket number off hand). The response I got (nearly a year later:) was that it was, at best, a documentation bug, and not something that needed to be fixed. I disagreed quite strongly with this.

I don't have a problem with the two versions doing something different if it was clearly documented, then you could make your choice of which form to use dependant upon which result fits your needs.

My concern was more that the result of the lvalue version produces a result that is neither one thing nor t'other.

$s='foobar';
print '4-arg: ', substr( $t=$s,1,4,$_)
   , ' 3-arg: ', substr($t=$s,1,4)=$_ 
   for qw[ a ab abc abcd abcde abcdef ];

4-arg:  ooba  3-arg:  ar
4-arg:  ooba  3-arg:  abr
4-arg:  ooba  3-arg:  abcr
4-arg:  ooba  3-arg:  abcd
4-arg:  ooba  3-arg:  abcd
4-arg:  ooba  3-arg:  abcd
[download]

As you can see, the 4-arg results are consistent, but the 3-arg version varies wildly.

1. If the value being assigned is shorter than the section being replaced, the result is the length of the assignment, but its value is
   1. the value of the assigned value
   2. PLUS a random character from the unaffected part of the original value...in this case the last character. </li
2. If the value being assigned is the same length as the section being replaced then the result is the value being assigned.
3. If the value being assigned is longer than the section being replaced, then the result is the first N characters of the value being assigned, where N is the length of the section being replaced.

No one is ever going to convince me that simply documenting this behaviour will make that logical, predicable or useful......still. "The problem space is a mess" :)